TDWI Upside - Where Data Means Business

Mid-Life Crisis? The Analytics Database Appliance at Middle Age

The analytics appliance is fifteen years old, but tech marketers and some industry experts say there's still plenty of demand for MPP database appliances.

In marketing, what's hot is often what's new, but even older products can be valuable. After all, although they're no longer sales leaders, development expenses and other costs have already been recouped, meaning more of each additional sale goes straight to the bottom line.

The analytics database appliance is fully 15 years into its evolution, and despite its age, tech marketers and some industry experts say there's still plenty of demand for massively parallel processing (MPP) database appliances.

One Active Product: IBM's Netezza

Former IBMer Lamont Lockwood, now a technical marketing specialist with Texas-based system integrator Reverent Technologies, describes IBM's Netezza MPP appliance as "a diamond in the rough." Lockwood says Big Blue's 16TB "Mini Mako" appliance bundles several IBM software options -- including data integration and business intelligence (BI) offerings -- along with the Netezza appliance hardware. Lockwood says Reverent is marketing the Netezza Mako systems to customers.

"If you buy any Mini Mako Netezza box, you get a little bit of IBM InfoSphere DataStage [ETL], a little bit of Cognos BI, a little bit of InfoSphere Streams, and a little bit of BigInsights, which is IBM'sHadoop [distribution]," Lockwood says.

"The only limitation is that it has to target the Netezza box either as a target or a source. That's a pretty reasonable limitation. If you run workloads that go into or out of Netezza, there's no limit."

Targeting Late-Adopting Customers

"When you hit the late-adopters' stage [of a product's maturity curve], there are a lot more customers. At the beginning, there are the [customers] you might think of as 'hobbyists.' There're only a few of these, and they're not cost-sensitive," he says, citing the long lead-up to profitability for analytics database sales.

In the 2000s, he points out, Netezza and its imitators -- DATAllegro (acquired by Microsoft), Dataupia, Greenplum, and even IBM itself -- were scrapping over several dozen bleeding-edge and early-adopter customers.

"Then you start ramping up. You get to the early adopters' stage, then you get to the peak, and there are a lot of competitors and a lot of customers but there are more deals, too. When you get to that late stage of the market, there are always lots of late adopters."

Steven Sarsfield, product marketing leader with Hewlett-Packard Enterprise Vertica, shares a similar perspective. There's still plenty of demand for MPP appliances such as Vertica, he argues. If anything, demand is actually increasing in proportion to the increasing size and complexity of data volumes and analytics workloads, respectively.

"Data volumes are growing and growing and the technology that people have in place is too expensive to handle this [growth]; it doesn't have the capacity or the capability. A lot of people are looking at alternative technologies now," he says.

Different Appliances for Different Purposes

"Hadoop is interesting in that it's a cost-effective way to store data, but for doing [decision-support] analytics, for doing advanced analytics, for doing concurrency, we have customers coming to us because Hadoop can't scale," Sarsfield continues. "We've done some benchmarks around how many concurrent queries a lot of these solutions can run. Usually it's in the single-digit numbers. If you have a database such as Vertica or even some of our competitors, you can hit thousands."

There's another consideration, too, says Sarsfield. General-purpose parallel processing platforms such as Hadoop or the Spark cluster computing framework address fundamentally different use cases than the MPP database appliance. "These aren't the same buyers [for each technology]. The guy who's using image-recognition analytics in Spark to do license-plate recognition is typically not a database guy. He's typically not buying Teradata or Vertica," he argues.

The larger point is that the data warehouse -- or data warehouse-like architecture -- isn't going anywhere, Sarsfield says. Many organizations are running out of gas on their existing -- single-node, SMP, row-based -- data warehouse systems.

"I think there are people who are still stuck with their typical data warehouse, their row-based data warehouse. They're trying to build indices; they're trying to build materialized views or cubes, and they're working very hard to get this thing fast enough so they can deliver analytics fast enough for their users," he says.

Believe it or not, he argues, the database appliance is still an unknown to some of these customers. "Is the fact that something like Vertica exists a surprise to these customers? Yes, I really think it is. It's hard for them to believe that something runs this fast, too," he says.

Not All Trends Positive for Appliances

Steve Dine, managing partner and founder of data management consultancy Datasource Consulting, agrees there's something to this. Based on what he sees in his customer engagements, there's still a good bit of demand for analytics appliances.

He appends a qualifier, however. "There are two trends that I think are hurting demand for Netezza and Vertica: [these] are mixed workloads and cloud. More organizations are looking to the cloud so they'll need to find a way to become more relevant in that market. I know that IBM has been pushing their cloud, but I'm not sure what their [cloud] offering [is] for Netezza," says Dine, who notes that third-party providers, such as Gigacom, do offer Netezza-as-a-service.

When it comes to support for mixed OLTP and analytics workloads, MPP analytics databases are less compelling than alternatives, he argues. "Exadata seems to be winning that [mixed-workload] game. I think HANA will also become a player in that space if SAP can broaden its market outside of SAP customers," says Dine.

He points out that both Exadata and HANA are combined OLTP- and analytics-processing platforms. "Companies want to host as few servers on site as possible. If they need separate solutions for operational and analytics workloads, then they want to look to the cloud."

Future of the Market

The database appliance isn't a new category, but Dine thinks things are only just starting to get interesting in this space. "As the cost of memory decreases and networks get faster, it will likely open the door to new database entrants that will challenge the incumbents," he argues.

It's possible, then, that MPP appliance price points could drop even more -- and that, in addition to dropping prices, IBM and other vendors could offer even sweeter software bundling deals, too. It's likewise possible Big Blue and other vendors could cede the low-end of the market -- and, with it, the MPP appliance segment -- to the next-gen database entrants Dine anticipates.

In any case, the analytics appliance market bears watching.

TDWI Membership

Accelerate Your Projects,
and Your Career

TDWI Members have access to exclusive research reports, publications, communities and training.

Individual, Student, & Team memberships available.